Pattern Catalog

The CREATE funnel is the foundational diagnostic model for every flow in a product. Each stage represents a cognitive gate the user must pass through before acting. If any gate fails, the action does not happen. For AI agents, the funnel provides a diagnostic framework: when a user stalls, identify which gate is blocked and intervene there. When used correctly, CREATE transforms product design from guesswork into behavioral engineering.

BJ Fogg's simplest and most practical diagnostic tool: if a behavior isn't happening, one of the three elements is missing or too weak.

Surfaces real evidence of other users' positive behavior to reduce uncertainty and accelerate action through the power of the crowd.

The speed bump for critical decisions. While most UX design aims to reduce friction, Strategic Friction deliberately introduces it when the stakes are high enough to warrant careful deliberation. Essential in deployment and financial contexts where a fast, unconsidered action can have material operational or monetary consequences. For AI agents, this is a mandatory checkpoint -- the agent cannot bypass friction on high-risk actions regardless of autonomy level.

A visible interface control that lets the user choose how much the AI agent may act on their behalf. Three positions: Suggest Only (agent drafts, user decides), Confirm Execution (agent proposes a plan, user approves), and Autonomous (agent acts, user reviews after). This is the central mechanism for maintaining human agency in an agent-first product.

BJ Fogg's proven recipe for automatic behavior: Anchor (existing routine) + Tiny Action (under 2 seconds) + Celebration (instant positive emotion).

The Default Effect leverages the human tendency to accept pre-selected options, making the desired action the easiest possible action.

The If-Then planning technique that creates pre-decided links between specific cues and specific actions, dramatically increasing follow-through.

Deliberately highlights what users lose by not acting, leveraging Prospect Theory to create urgency without manipulation.

Mechanisms that lock in future behavior by raising the cost of deviation, bridging the gap between today's motivation and tomorrow's action.

From Wendel's Designing for Behavior Change: every target action falls into one of three strategy buckets. Cheat automates or defaults the action so the user barely thinks. Habits turn the action into an automatic routine using Fogg's B=MAT model. Conscious Action supports high-stakes deliberation with full information and strategic friction. Choosing the right strategy before designing any UI makes the rest of the work dramatically easier.

Visual and textual signals that communicate how confident the AI agent is in its output. Prevents the single most damaging failure mode in AI-assisted workflows: presenting a guess as a fact. Confidence signals use color, iconography, percentage indicators, and calibrated language to help users decide how much scrutiny to apply.

Effort and persistence increase as perceived distance to the finish line shrinks. Visually shorten remaining progress to sustain momentum.

Delivers outcomes on random or semi-random schedules to sustain engagement through anticipation and novelty.

Deliberately leaves tasks incomplete or visually open to create psychological pull that drives users back to close the loop.

Decision time increases logarithmically with options. Limit visible choices to 3-7, prioritize the best option, use progressive disclosure.

Design the emotional high point and closing moment of every user journey to create lasting positive memory.

Wendel's three-step process for making any action easier. First, Tailor it to the specific actor and context. Then Simplify by removing unnecessary steps and decisions. Finally, Make it Easy by reducing physical and cognitive effort through defaults, templates, and one-click paths. This pattern is applied before UI design begins -- it shapes what the interface needs to support.

A living document that maps how a user moves from first encounter to habitual use of a feature, using the CREATE funnel as the diagnostic backbone. Written before any UI is designed, it serves as the contract between design intent and implementation. For AI agents, the Behavioral Plan defines success criteria: the agent's job is to move the user through the plan's stages, not just complete a task.

Make every interactive element instantly understandable through clear affordances (what it does) and signifiers (how it looks).

Prevent errors where possible, catch them early, and make recovery effortless with one-click undo and kind messaging.

Close the loop between user action and system response with immediate, clear, and appropriate feedback.

Deliberately creates a small information void right before a desired action, leveraging the brain's drive to resolve uncertainty.

Create the formula 'After [current habit], I will [new habit]' to bootstrap new behaviors onto existing reliable routines.

Ask for the smallest possible yes that still moves the user forward, then use momentum to build larger behaviors.

Deliberately design how options are structured, ordered, grouped, and framed to guide users toward better decisions.

Gently lower the barrier to beneficial change while preserving the safety of the current option through reversible, low-friction migration.

Bridge the gap between now and later by making future benefits feel immediate and tangible through visualization and instant micro-rewards.

Make every meaningful change undoable with one click, giving users psychological safety to experiment and trust the system.

Surface, check, and gently correct the user's understanding of the system in real time to eliminate confusion.

Know exactly when to pause automation and yield to the human at the precise point where human judgment adds the most value.

Make every agent action transparent by showing what was done, why, and what data was used -- eliminating the black box feeling.

Dynamically signal exactly how much users should trust each agent output using visual weight, language, and evidence linkage.

Dynamically adjust agent autonomy based on real-time signals from task risk, user trust, and confidence -- without user micromanagement.

Force the agent to show its work before acting -- outline every step, data source, and expected outcome for informed consent.

A living history log that makes every agent suggestion, decision, and execution fully traceable and reversible.

Show only essential elements by default, reveal advanced options when contextually relevant, and provide a clear way to expand or collapse.